fuzz: add fuzztests that try to correct bech32 and codex32 errors

The codex32 test will more thoroughly exercise the algebra, since there
we can correct up to 4 errors. The bech32 test on the other hand should
work without an allocator (though to exercise this you need to manually
edit fuzz/Cargo.toml to disable the alloc feature -- this is

rust-lang/cargo#2980

which has been open for 10 years and counting..)

Author: apoelstra

fuzz/Cargo.toml

@@ -21,6 +21,14 @@ members = ["."]
 name = "berlekamp_massey"
 path = "fuzz_targets/berlekamp_massey.rs"
 
+[[bin]]
+name = "correct_bech32"
+path = "fuzz_targets/correct_bech32.rs"
+
+[[bin]]
+name = "correct_codex32"
+path = "fuzz_targets/correct_codex32.rs"
+
 [[bin]]
 name = "decode_rnd"
 path = "fuzz_targets/decode_rnd.rs"

fuzz/fuzz_targets/correct_bech32.rs

@@ -0,0 +1,93 @@
+use std::collections::HashMap;
+
+use bech32::primitives::correction::CorrectableError as _;
+use bech32::primitives::decode::CheckedHrpstring;
+use bech32::{Bech32, Fe32};
+use honggfuzz::fuzz;
+
+// coinbase output of block 862290
+static CORRECT: &[u8; 62] = b"bc1qwzrryqr3ja8w7hnja2spmkgfdcgvqwp5swz4af4ngsjecfz0w0pqud7k38";
+
+fn do_test(data: &[u8]) {
+    if data.is_empty() || data.len() % 2 == 1 {
+        return;
+    }
+
+    // Start with a correct string
+    let mut hrpstring = *CORRECT;
+    // ..then mangle it
+    let mut errors = HashMap::with_capacity(data.len() / 2);
+    for sl in data.chunks_exact(2) {
+        let idx = usize::from(sl[0]);
+        if idx >= CORRECT.len() - 3 {
+            return;
+        }
+        let offs = match Fe32::try_from(sl[1]) {
+            Ok(Fe32::Q) => return,
+            Ok(fe) => fe,
+            Err(_) => return,
+        };
+
+        hrpstring[idx + 3] =
+            (Fe32::from_char(hrpstring[idx + 3].into()).unwrap() + offs).to_char() as u8;
+        if errors.insert(idx + 3, offs).is_some() {
+            return;
+        }
+    }
+
+    let s = unsafe { core::str::from_utf8_unchecked(&hrpstring) };
+    /*
+    println!("{}", unsafe { core::str::from_utf8_unchecked(CORRECT) });
+    println!("{}", s);
+    */
+    let corrections = CheckedHrpstring::new::<Bech32>(s)
+        .unwrap_err()
+        .correction_context::<Bech32>()
+        .unwrap()
+        .bch_errors();
+
+    if errors.len() <= 4 {
+        for (idx, fe) in corrections.unwrap() {
+            let idx = s.len() - idx - 1;
+            //println!("Errors: {:?}", errors);
+            //println!("Remove: {} {}", idx, fe);
+            assert_eq!(errors.remove(&idx), Some(fe));
+        }
+        assert_eq!(errors.len(), 0);
+    }
+}
+
+fn main() {
+    loop {
+        fuzz!(|data| {
+            do_test(data);
+        });
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    fn extend_vec_from_hex(hex: &str, out: &mut Vec<u8>) {
+        let mut b = 0;
+        for (idx, c) in hex.as_bytes().iter().filter(|&&c| c != b'\n').enumerate() {
+            b <<= 4;
+            match *c {
+                b'A'..=b'F' => b |= c - b'A' + 10,
+                b'a'..=b'f' => b |= c - b'a' + 10,
+                b'0'..=b'9' => b |= c - b'0',
+                _ => panic!("Bad hex"),
+            }
+            if (idx & 1) == 1 {
+                out.push(b);
+                b = 0;
+            }
+        }
+    }
+
+    #[test]
+    fn duplicate_crash() {
+        let mut a = Vec::new();
+        extend_vec_from_hex("", &mut a);
+        super::do_test(&a);
+    }
+}

fuzz/fuzz_targets/correct_codex32.rs

@@ -0,0 +1,118 @@
+use std::collections::HashMap;
+
+use bech32::primitives::correction::CorrectableError as _;
+use bech32::primitives::decode::CheckedHrpstring;
+use bech32::{Checksum, Fe1024, Fe32};
+use honggfuzz::fuzz;
+
+/// The codex32 checksum algorithm, defined in BIP-93.
+///
+/// Used in this fuzztest because it can correct up to 4 errors, vs bech32 which
+/// can correct only 1. Should exhibit more interesting behavior.
+#[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub enum Codex32 {}
+
+impl Checksum for Codex32 {
+    type MidstateRepr = u128;
+    type CorrectionField = Fe1024;
+    const ROOT_GENERATOR: Self::CorrectionField = Fe1024::new([Fe32::_9, Fe32::_9]);
+    const ROOT_EXPONENTS: core::ops::RangeInclusive<usize> = 9..=16;
+
+    const CHECKSUM_LENGTH: usize = 13;
+    const CODE_LENGTH: usize = 93;
+    // Copied from BIP-93
+    const GENERATOR_SH: [u128; 5] = [
+        0x19dc500ce73fde210,
+        0x1bfae00def77fe529,
+        0x1fbd920fffe7bee52,
+        0x1739640bdeee3fdad,
+        0x07729a039cfc75f5a,
+    ];
+    const TARGET_RESIDUE: u128 = 0x10ce0795c2fd1e62a;
+}
+
+static CORRECT: &[u8; 48] = b"ms10testsxxxxxxxxxxxxxxxxxxxxxxxxxx4nzvca9cmczlw";
+
+fn do_test(data: &[u8]) {
+    if data.is_empty() || data.len() % 2 == 1 {
+        return;
+    }
+
+    // Start with a correct string
+    let mut hrpstring = *CORRECT;
+    // ..then mangle it
+    let mut errors = HashMap::with_capacity(data.len() / 2);
+    for sl in data.chunks_exact(2) {
+        let idx = usize::from(sl[0]);
+        if idx >= CORRECT.len() - 3 {
+            return;
+        }
+        let offs = match Fe32::try_from(sl[1]) {
+            Ok(Fe32::Q) => return,
+            Ok(fe) => fe,
+            Err(_) => return,
+        };
+
+        hrpstring[idx + 3] =
+            (Fe32::from_char(hrpstring[idx + 3].into()).unwrap() + offs).to_char() as u8;
+        if errors.insert(idx + 3, offs).is_some() {
+            return;
+        }
+    }
+
+    let s = unsafe { core::str::from_utf8_unchecked(&hrpstring) };
+    /*
+    println!("{}", unsafe { core::str::from_utf8_unchecked(CORRECT) });
+    println!("{}", s);
+    */
+    let corrections = CheckedHrpstring::new::<Codex32>(s)
+        .unwrap_err()
+        .correction_context::<Codex32>()
+        .unwrap()
+        .bch_errors();
+
+    if errors.len() <= 4 {
+        for (idx, fe) in corrections.unwrap() {
+            let idx = s.len() - idx - 1;
+            //println!("Errors: {:?}", errors);
+            //println!("Remove: {} {}", idx, fe);
+            assert_eq!(errors.remove(&idx), Some(fe));
+        }
+        assert_eq!(errors.len(), 0);
+    }
+}
+
+fn main() {
+    loop {
+        fuzz!(|data| {
+            do_test(data);
+        });
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    fn extend_vec_from_hex(hex: &str, out: &mut Vec<u8>) {
+        let mut b = 0;
+        for (idx, c) in hex.as_bytes().iter().filter(|&&c| c != b'\n').enumerate() {
+            b <<= 4;
+            match *c {
+                b'A'..=b'F' => b |= c - b'A' + 10,
+                b'a'..=b'f' => b |= c - b'a' + 10,
+                b'0'..=b'9' => b |= c - b'0',
+                _ => panic!("Bad hex"),
+            }
+            if (idx & 1) == 1 {
+                out.push(b);
+                b = 0;
+            }
+        }
+    }
+
+    #[test]
+    fn duplicate_crash() {
+        let mut a = Vec::new();
+        extend_vec_from_hex("", &mut a);
+        super::do_test(&a);
+    }
+}